A historian, a biologist and a physicist walk into a bar. The barista, a part-time epistemologist (like most), asks each of them: Can you please describe what it means for you to know something? Nothing very funny ensues[1] -- but my hope is to eventually figure out the punchline.
Even restricting ourselves to modern academic cultures and the most popular neurotypes, there are clearly different ways of creating and receiving knowledge that are not very transparent to each other. Often, what other research traditions investigate and produce appears to be, at best, boring or shallow or confusing, and at worst, not knowledge at all (I am sure you can think of a few).
Yet I am an optimist: I believe that the most crucial divergences and misunderstandings lie along a limited number of universal dimensions, and that we could someday hope to draw each research tradition's contour on an epistemic map -- a navigation tool to make sense of, and ideally taste the beauty and joy of, other ways of knowing or "thinking styles".
Now, many researchers from many disciplines have come up with their own typologies of sciences, or of scientists, or of knowledge. Some look like they are talking about roughly the same things in various combinations and rotations, some appear entirely orthogonal, and some are, quite frankly, not too interesting to me (like those that worry a lot about what makes knowledge justified and true) though I hope to learn how to enjoy them as well. In the spirit of adding one more grand theory of everything to the list, I have been working on my own typology, with the aim of:
coming up with clever ways of revealing implicitly held thinking styles: questionnaires, interviews, original formats for structured dialogue or presentations, and games,
inflicting these formats upon people from many fields,
building epistemic maps from their answers and reactions,
and hopefully using these maps to facilitate someone's understanding of their own discipline as well as of others.
To start this process, I first wanted to gather the broadest-while-still-interesting-to-me summary of what various prior epistemic typologies have tried to capture so far - including all the ones I know, but downplaying the intricacies of those I care less about. This post is a first report on this effort.
Disclaimer: You might be expecting a literature review, but nowadays that feels more and more like an LLM's job, so I will instead focus on offering you my own original cognitive output, without much of the ancestor worship that makes proper scholarship. Let me just condense here the ritual name-dropping, so you can assume I've read and thought about the following at least: Pepper, Popper, Lakatos, Lakoff, Hesse, Hempel, Hacking, Habermas, Kuhn, Kitcher, Kimble, Crombie, Quine, van Fraassen, Bechtel, Black.
In brief: a few orthogonal dimensions
After reading a bunch of epistemic typologies, it seemed to me that they were dealing with at least four more-or-less unrelated questions:
What makes a piece of knowledge (question or answer) valid, true or legitimate from the point of view of an individual or a community.
What basic cognitive abilities are involved in its manipulation.
What makes it interesting to someone.
What kind of move it represents in the space of knowledge.
The third and fourth questions are the most interesting to me, while the first and second are the most explored in the literature (as two rather distinct fields), hence my feeling that I have something to contribute to this whole game. My intuition as to why I care less about the former is that they feel more rigid, imposed by social norms or by personal abilities, and thus less likely to allow exploration and self-modification.
The (rigid?) dimensions that I care a bit less about
(J) Epistemic justification: When does a claim count as true?
I start with this one, and I will say very little about it, because truth value and justification have long been the main question of epistemology, from logical positivists to postmodernists to various other religious groups (including Talmudists, Buddhists, Mimamsakas, etc).
The classic topics are:
Which modes and forms of knowing are trustworthy: various sensory organs, analytical reasoning, repeated experiment, syllogism, dream, word of mouth, etc.
Which people, communities or entities are valid sources: myself only, certain human authorities (e.g. sage kings, the scientific community), certain non-human authorities (e.g. the Bible or Vedas), etc.
The usual approach is granting some value to most or all these modes and sources, but ranking them so we know which one to trust if two come in conflict.
There is something deeply ideological about any such posture: if you grew up believing that the Vedas are the source code of the universe and are absolutely right even when contradicted by every other kind of evidence, or that your direct experience of things overcomes any level of authority and consensus from so-called experts, then thinking differently becomes a matter of religious conversion. I will leave this to whoever enjoys the proselytizing business more than I do.
(C) Cognitive tools: How does it work "concretely" in one's mind?
The idea in brief is that, while we like to pretend that understanding is a purely logical process operating in the world of abstract concepts, the reality is that we only manage to reason about something by recasting it in terms of one of our basic cognitive abilities (senses, language, etc.) and manipulating it in some way. The issue is that we often do not realize it ourselves: when I intuit the solution to a physics problem, am I pattern-matching to an auditory sequence of "minus b plus or minus square root of ...", or visualizing shapes on the blackboard of my mind, or feeling things moving and bumping into each other?
Particular modules that can be used to think about other things
Words, symbolic manipulation, drawing on grammatical skills
Theory of mind: ability to infer beliefs and drives and to reason about agency
General processing abilities and styles
Recall of learned ideas versus generation of new ideas on the spot
Tempo: slow versus fast processing, instantaneous versus delayed response
Ability to access the self, sensations...
Deep or broad attention (context insensitive vs sensitive, background absent or present)
This is presumably a very useful thing to know. I was told that learning to observe how I actually think in these concrete terms, and talking about them explicitly, could dramatically help others understand what I am thinking and how. I confess that I have not tried it seriously.
Why I am a bit less interested in this, besides how it correlates with the dimensions below:
This is the turf of much research on cognition and pedagogy, and it is a very fine turf but also a well established one, and it is not going to be very easy to improve.
This feels more like an issue of personal ability, not very easy to change (though sometimes worth it), and often tied to individual "styles" within a field rather than different fields altogether: there might not be many geometers without visual intuition, but I have met theoretical physicists and ecologists who think in symbols and others who think in shapes. It impacts how you solve problems, and even which kind of theoretical discourse you like the most, but I believe these various styles are ultimately able to contribute to the same questions, whereas the dimensions below may yield entirely different questions.
The (flexible?) dimensions that I care a bit more about
(Q) Qualities and aesthetics: What makes a piece of knowledge salient and interesting?
Some of the properties listed below can perhaps be used to argue about truth value and justification, others cannot, but all play into what makes a piece of knowledge feel interesting and beautiful.
NB: There are aspects that I feel no need to mention because I intuit they will always increase or decrease this sense of salience regardless of thinking style, e.g. whether it is generative of new ideas (any insight is more beautiful when it leads to more insights). I want to focus here on aspects for which researchers may have different preferences.
Legibility or ease of access: is this piece of knowledge...
Shared with others, amenable to (almost) lossless translation[2]
Private but explicit: I can notice, access and manipulate it even if I cannot share it
Tacit: I can only "stand in it" in order to know (or act upon) other things
Relation to other pieces of knowledge: breadth, contrast
General vs particular: connecting many objects or delving deeply into one, showing how things are similar or different
Surprise and novelty vs accumulation (few or many other claims going in the same direction)
Consistency: (in)tolerance for contradictory claims
Precision: clarity of contrast and boundary with other claims
Close-ended vs open-ended: what matters more is the answer (solvable problem) or the question (inexhaustible mystery)
Observation versus action (what changes in the process - the map, the territory, the cartographer?)
in the means of knowing: to which extent acquiring knowledge involves acting on, or interacting with, what is being known
Description, measurement
Experiment
Intersubjective encounter
in the purpose of knowing: once acquired, what this can be used for (is this knowledge positive or negative, constructive or critical, etc?)
Extending knowledge (understanding or predicting)
Transforming knowledge into action (building or controlling)
Living with, relating to something
Reflexivity, seeing or changing oneself and/or other knowers
Intuitively, I feel that these features combine to form distinct ways of enjoying knowledge. Just a few examples in clichés:
A collector likes exceptions to every rule, she seeks to learn about dizzying numbers of entities and the subtle fingerprints that makes each of them radically unique, finding stable difference where one might see uniform chaos. A theoretician also wishes to encompass dizzying numbers of examples, but memorize none of them, and instead find similarity where there seems to be difference.
A philosopher and a mystic may both want to sit with a question rather than get an answer, yet the philosopher will insist on exacting precision and translatability and knowledge expansion, and the mystic will find joy in holding something private and diffuse and transformative of the self.
An artist may want to change the world with questions, an engineer with answers. And so on, and so forth. There may be a more compact theory than the complete combinatorics of the above[3], but many combinations seem plausible and interpretable.
(M) Moves in knowledge space: How does one use things to learn about other things?
This is the dimension that has most to do with the actual content of knowledge, or at least with broad kinds of content. The issue here is not what makes a question or claim feel interesting or true, but how it generates knowledge, and under which conditions it fails to do so (and thus feels empty and nonsensical, e.g. Borges's Celestial Emporium).
The intuition here is that knowledge is primarily motion: it gives us relative paths (GPS directions) in the space of mental objects, telling us how to jump from one to another, using the former to understand the latter by transferring intuitions and insights. The point is that the move becomes meaningless if one does not start from the right location, if the correct source is either not provided or not intuitive to us; thus, an advanced argument in economics or theology can seem absurd unless one has internalized a lot of other prevailing arguments and come to believe in their epistemic value and power.
These kinds of moves between objects of knowledge can be called model, deduction or induction, metaphor or analogy, even explanation (in the sense of Y explaining X). I am trying to embrace something more general than any of these classic concepts of epistemology, to understand what they all share:
Identifying the particular source and target
Micro-level (individual moves): which object or phenomenon or concept am I using to think about another, e.g.
a formula to think about a physical phenomenon
atoms to think about a gas
a simplified model of human psychology to think about a song's emotional impact
Meso-level: Kuhnian exemplars that structure a field, e.g. springs and balls for almost all of physics
Macro-level: Using a field as a template to think about another e.g. thinking of dietetics by analogy to biomedical research and its methodology (RCT etc.)
General characteristics of that move
Abstract vs concrete knowledge = moving between distant or nearby objects (e.g. from formula to physical object, or between two lab rodents)
What kind of relation exists between the source and the target (I have a whole article about this so I won't enter into details here, at least for now)
Comparison: a source is some other entity that has no a priori relation to the target, letting relevant features emerge (juxtaposetwo entities, what emerges?)
Composition: potential sources are parts of the target, constituents or properties (transplantthe target in a new setting, what moves with it?)
Context: potential sources are settings and systems in which the target plays a role (substitutesomething else for the target, what stays the same?)
Contingency: potential sources are accidents that have happened or can happen to the target (re-runthe same target in different circumstances, what changes?)
What do I get out of this?
Even as a bare sketch, laying out the dimensions above has made it easy for me to slot any new typology of science I encounter (so far) into some sort of cohesive coordinate space - and what initially confused me about these typologies can often be analyzed as them mixing several of the axes inconsistently. Still, reviews and classifications for their own sake are perhaps a little sterile, and my aim is rather to push for new research:
Trying to better understand the dimensions Q and M that are far less clearly explored in the literature.
As I said in the introduction, use this to actually interrogate researchers and understand them better (and ideally find tricks and skills to help them understand themselves and others)
In the process, figure out if the dimensions and sub-dimensions listed above are sufficiently exhaustive or missing something that feels essential in practice, independent or correlated in obvious or surprising ways, etc.
In the spirit of fast iteration, I make this public and may at some point edit it heavily or write a new version altogether, so I welcome any thoughts and comments!
If you are like me, you may have a hard time imagining that someone might not want this. But there might be fields that yield no knowledge of this kind except simply encyclopedic and factual description, e.g. "Bob the cat has long hair." which is indeed quite widely shareable but not very enticing.
I feel that one potential organizing prism that may be popular in these parts would be the metaphor of a predictive machine in a collective. - finesse = shrinking the residuals to zero on a small part of the model - compression and generalization = shrinking the model - progress, trust, consistency = sharing the model across knowers and tasks - mystery = refining the gap/error detector itself - empowerment = enactive prediction - etc. This still feels way too open-ended at the moment, but I guess it's a start.
A historian, a biologist and a physicist walk into a bar.
The barista, a part-time epistemologist (like most), asks each of them: Can you please describe what it means for you to know something?
Nothing very funny ensues[1] -- but my hope is to eventually figure out the punchline.
Even restricting ourselves to modern academic cultures and the most popular neurotypes, there are clearly different ways of creating and receiving knowledge that are not very transparent to each other. Often, what other research traditions investigate and produce appears to be, at best, boring or shallow or confusing, and at worst, not knowledge at all (I am sure you can think of a few).
Yet I am an optimist: I believe that the most crucial divergences and misunderstandings lie along a limited number of universal dimensions, and that we could someday hope to draw each research tradition's contour on an epistemic map -- a navigation tool to make sense of, and ideally taste the beauty and joy of, other ways of knowing or "thinking styles".
Now, many researchers from many disciplines have come up with their own typologies of sciences, or of scientists, or of knowledge. Some look like they are talking about roughly the same things in various combinations and rotations, some appear entirely orthogonal, and some are, quite frankly, not too interesting to me (like those that worry a lot about what makes knowledge justified and true) though I hope to learn how to enjoy them as well. In the spirit of adding one more grand theory of everything to the list, I have been working on my own typology, with the aim of:
To start this process, I first wanted to gather the broadest-while-still-interesting-to-me summary of what various prior epistemic typologies have tried to capture so far - including all the ones I know, but downplaying the intricacies of those I care less about. This post is a first report on this effort.
Disclaimer: You might be expecting a literature review, but nowadays that feels more and more like an LLM's job, so I will instead focus on offering you my own original cognitive output, without much of the ancestor worship that makes proper scholarship. Let me just condense here the ritual name-dropping, so you can assume I've read and thought about the following at least: Pepper, Popper, Lakatos, Lakoff, Hesse, Hempel, Hacking, Habermas, Kuhn, Kitcher, Kimble, Crombie, Quine, van Fraassen, Bechtel, Black.
In brief: a few orthogonal dimensions
After reading a bunch of epistemic typologies, it seemed to me that they were dealing with at least four more-or-less unrelated questions:
The third and fourth questions are the most interesting to me, while the first and second are the most explored in the literature (as two rather distinct fields), hence my feeling that I have something to contribute to this whole game. My intuition as to why I care less about the former is that they feel more rigid, imposed by social norms or by personal abilities, and thus less likely to allow exploration and self-modification.
The (rigid?) dimensions that I care a bit less about
(J) Epistemic justification: When does a claim count as true?
I start with this one, and I will say very little about it, because truth value and justification have long been the main question of epistemology, from logical positivists to postmodernists to various other religious groups (including Talmudists, Buddhists, Mimamsakas, etc).
The classic topics are:
The usual approach is granting some value to most or all these modes and sources, but ranking them so we know which one to trust if two come in conflict.
There is something deeply ideological about any such posture: if you grew up believing that the Vedas are the source code of the universe and are absolutely right even when contradicted by every other kind of evidence, or that your direct experience of things overcomes any level of authority and consensus from so-called experts, then thinking differently becomes a matter of religious conversion. I will leave this to whoever enjoys the proselytizing business more than I do.
(C) Cognitive tools: How does it work "concretely" in one's mind?
The idea in brief is that, while we like to pretend that understanding is a purely logical process operating in the world of abstract concepts, the reality is that we only manage to reason about something by recasting it in terms of one of our basic cognitive abilities (senses, language, etc.) and manipulating it in some way. The issue is that we often do not realize it ourselves: when I intuit the solution to a physics problem, am I pattern-matching to an auditory sequence of "minus b plus or minus square root of ...", or visualizing shapes on the blackboard of my mind, or feeling things moving and bumping into each other?
Particular modules that can be used to think about other things
General processing abilities and styles
This is presumably a very useful thing to know. I was told that learning to observe how I actually think in these concrete terms, and talking about them explicitly, could dramatically help others understand what I am thinking and how. I confess that I have not tried it seriously.
Why I am a bit less interested in this, besides how it correlates with the dimensions below:
The (flexible?) dimensions that I care a bit more about
(Q) Qualities and aesthetics: What makes a piece of knowledge salient and interesting?
Some of the properties listed below can perhaps be used to argue about truth value and justification, others cannot, but all play into what makes a piece of knowledge feel interesting and beautiful.
NB: There are aspects that I feel no need to mention because I intuit they will always increase or decrease this sense of salience regardless of thinking style, e.g. whether it is generative of new ideas (any insight is more beautiful when it leads to more insights). I want to focus here on aspects for which researchers may have different preferences.
Intuitively, I feel that these features combine to form distinct ways of enjoying knowledge. Just a few examples in clichés:
A collector likes exceptions to every rule, she seeks to learn about dizzying numbers of entities and the subtle fingerprints that makes each of them radically unique, finding stable difference where one might see uniform chaos. A theoretician also wishes to encompass dizzying numbers of examples, but memorize none of them, and instead find similarity where there seems to be difference.
A philosopher and a mystic may both want to sit with a question rather than get an answer, yet the philosopher will insist on exacting precision and translatability and knowledge expansion, and the mystic will find joy in holding something private and diffuse and transformative of the self.
An artist may want to change the world with questions, an engineer with answers. And so on, and so forth. There may be a more compact theory than the complete combinatorics of the above[3], but many combinations seem plausible and interpretable.
(M) Moves in knowledge space: How does one use things to learn about other things?
This is the dimension that has most to do with the actual content of knowledge, or at least with broad kinds of content. The issue here is not what makes a question or claim feel interesting or true, but how it generates knowledge, and under which conditions it fails to do so (and thus feels empty and nonsensical, e.g. Borges's Celestial Emporium).
The intuition here is that knowledge is primarily motion: it gives us relative paths (GPS directions) in the space of mental objects, telling us how to jump from one to another, using the former to understand the latter by transferring intuitions and insights. The point is that the move becomes meaningless if one does not start from the right location, if the correct source is either not provided or not intuitive to us; thus, an advanced argument in economics or theology can seem absurd unless one has internalized a lot of other prevailing arguments and come to believe in their epistemic value and power.
These kinds of moves between objects of knowledge can be called model, deduction or induction, metaphor or analogy, even explanation (in the sense of Y explaining X). I am trying to embrace something more general than any of these classic concepts of epistemology, to understand what they all share:
What do I get out of this?
Even as a bare sketch, laying out the dimensions above has made it easy for me to slot any new typology of science I encounter (so far) into some sort of cohesive coordinate space - and what initially confused me about these typologies can often be analyzed as them mixing several of the axes inconsistently. Still, reviews and classifications for their own sake are perhaps a little sterile, and my aim is rather to push for new research:
In the spirit of fast iteration, I make this public and may at some point edit it heavily or write a new version altogether, so I welcome any thoughts and comments!
Unless My Dinner with Andre is your kind of joke.
If you are like me, you may have a hard time imagining that someone might not want this. But there might be fields that yield no knowledge of this kind except simply encyclopedic and factual description, e.g. "Bob the cat has long hair." which is indeed quite widely shareable but not very enticing.
I feel that one potential organizing prism that may be popular in these parts would be the metaphor of a predictive machine in a collective.
- finesse = shrinking the residuals to zero on a small part of the model
- compression and generalization = shrinking the model
- progress, trust, consistency = sharing the model across knowers and tasks
- mystery = refining the gap/error detector itself
- empowerment = enactive prediction
- etc.
This still feels way too open-ended at the moment, but I guess it's a start.